Glossary

Use this glossary to find definitions of Agent Hub terms and related AI Data Platform concepts that you encounter while using Agent Hub or reading this guide.

  • A2A - Agent2Agent (A2A) is an open protocol that enables AI agents to discover and interact with one another. An agent's A2A metadata enables the agent to be registered and discovered.
  • A2A agent card - An agent card contains metadata describing an agent, including its name, description, and skills. It supports agent discovery and interoperability.
  • A2UI - Agent to UI (A2UI) is a structured user interface protocol that enables an agent to return interactive user interface components in addition to text.
  • Agent flow - Agent flow is the code or visual definition that coordinates a model, instructions, and tools.
  • Agent Hub administrator - An Agent Hub administrator is a user who administers Agent Hub configuration and functionality.
  • AI agent - An AI agent is an AI resource designed to perform particular tasks or provide specialized knowledge and capabilities.
  • AI compute - AI compute is the runtime capacity attached to an agent flow for deployment.
  • Oracle AI Data Platform - AI Data Platform provides the environment where agents are built and deployed. Also provides platform services used by Agent Hub.
  • AI Registry - A central enterprise registry for registering and discovering supported AI resources developed in Oracle AI Data Platform or supported external environments. Registered resources can be made available for use through Agent Hub and other applications.
  • AI Resources - The AI capabilities that an organization makes available for use through Agent Hub, including agents, MCP servers, and tools.
  • Business user - A business user is a user who uses Agent Hub to interact with the enterprise AI resources available to them.
  • Checkpointer - A checkpointer provides conversation-scoped state storage used to retain multi-turn context.
  • Deployment - A deployment is a runnable TEST or PROD endpoint created from an agent flow.
  • Direct response - A direct response is a response that Agent Hub provides from its configured model without routing the request to a registered agent.
  • Enterprise AI - Enterprise AI refers to the AI capabilities that an organization develops, manages, and makes available for business use.
  • Guardrails - Guardrails are controls designed to protect AI interactions by detecting or preventing specified types of unwanted content or behavior.
  • Knowledge base - A knowledge base contains indexed documents that a retrieval-augmented generation tool can use to retrieve relevant evidence.
  • Large language model (LLM) - An AI model that can understand and generate natural language. Agent Hub uses a configured LLM to support orchestration and provide direct responses when a request doesn't require a specialized agent.
  • MCP server - A server that exposes tools through Model Context Protocol (MCP) so that agents and other supported functionality can discover and call them.
  • Model Context Protocol (MCP) - Model Context Protocol (MCP) is a standard interface through which an agent can discover and call external tools.
  • Orchestrator - An Agent Hub component that evaluates a user's request and determines how to handle it. The orchestrator can route requests to appropriate agents or tools, present agents for user selection, or provide a direct response using the configured large language model (LLM).
  • Prompt - A prompt is a question, instruction, or request that a business user enters in Agent Hub.
  • Registry - The Registry is the page in Agent Hub where business users browse and discover the agents, topics, and other supported AI resources available to them.
  • Retrieval-augmented generation (RAG) - Retrieval-augmented generation (RAG) is a technique in which a model generates a response using information retrieved from a knowledge base.
  • Role-based access control (RBAC) - Role-based access control (RBAC) controls access to functionality and resources based on roles and associated permissions.
  • Session variable - A session variable is client-provided runtime metadata delivered separately from the user's prompt.
  • SQL tool - A SQL tool provides a configured way for an agent to query registered data.
  • Thinking Activity - A panel that the user can display during a chat. The panel provides a high-level view of request processing, including processing status, elapsed time, and information about agents or tools invoked during a request.
  • Tool - A tool provides a capability that an agent or other supported Agent Hub functionality can use to perform a task or interact with another system.
  • Topic - A topic groups related agents and tools around a particular business area or purpose. When a user starts a chat from a topic, the topic narrows the set of agents and tools the orchestrator can use to process the request.
  • User role - A user role defines a set of permissions associated with a user's responsibilities or activities.